CHILDREN AND FAMILIES
EDUCATION AND THE ARTS
The RAND Corporation is a nonprofit institution that helps improve policy and
decisionmaking through research and analysis.
ENERGY AND ENVIRONMENT
HEALTH AND HEALTH CARE
INFRASTRUCTURE AND
TRANSPORTATION
This electronic document was made available from www.rand.org as a public service
of the RAND Corporation.
INTERNATIONAL AFFAIRS
LAW AND BUSINESS
NATIONAL SECURITY
Skip all front matter: Jump to Page 16
POPULATION AND AGING
PUBLIC SAFETY
SCIENCE AND TECHNOLOGY
TERRORISM AND
HOMELAND SECURITY
Support RAND
Browse Reports & Bookstore
�
Make a charitable contribution
�
For More Information
Visit RAND at www.rand.org
�
Explore RAND Labor and Population
�
View document details
�
Limited Electronic Distribution Rights
This document and trademark(s) contained herein are protected by law as indicated in a notice appearing later in this work.
This electronic representation of RAND intellectual property is provided for non-commercial use only. Unauthorized posting of
RAND electronic documents to a non-RAND Web site is prohibited. RAND electronic documents are protected under copyright
law. Permission is required from RAND to reproduce, or reuse in another form, any of our research documents for commercial
use. For information on reprint and linking permissions, please see RAND Permissions.
This product is part of the RAND Corporation technical report series. Reports may
include research findings on a specific topic that is limited in scope; present discussions
of the methodology employed in research; provide literature reviews, survey instruments, modeling exercises, guidelines for practitioners and research professionals, and
supporting documentation; or deliver preliminary findings. All RAND reports undergo rigorous peer review to ensure that they meet high standards for research quality
and objectivity.
TECHNIC A L REP O RT
Research Designs for Estimating
Induced Entry into the SSDI
Program Resulting from a
Benefit Offset
Nicole Maestas • Kathleen J. Mullen • Gema Zamarro
Sponsored by the Social Security Administration
LA BOR AND POPULATION
This research was funded by a contract from the Social Security Administration and was
undertaken within RAND Labor and Population.
The R AND Corporation is a nonprofit institution that helps improve policy and
decisionmaking through research and analysis. RAND’s publications do not necessarily
reflect the opinions of its research clients and sponsors.
R® is a registered trademark.
© Copyright 2010 RAND Corporation
Permission is given to duplicate this document for personal use only, as long as it
is unaltered and complete. Copies may not be duplicated for commercial purposes.
Unauthorized posting of RAND documents to a non-RAND website is prohibited. RAND
documents are protected under copyright law. For information on reprint and linking
permissions, please visit the RAND permissions page (http://www.rand.org/publications/
permissions.html).
Published 2010 by the RAND Corporation
1776 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138
1200 South Hayes Street, Arlington, VA 22202-5050
4570 Fifth Avenue, Suite 600, Pittsburgh, PA 15213-2665
RAND URL: http://www.rand.org
To order RAND documents or to obtain additional information, contact Distribution Services: Telephone: (310) 451-7002; Fax: (310) 451-6915; Email: order@rand.org
Summary To support the Social Security Administration (SSA) in fulfilling its legislative mandate
under the Ticket to Work Incentive and Work Incentives Improvement Act (P.L. 106-170;
the Ticket Act), this report has the principal aim of providing SSA with a set of research
design options for estimating induced entry effects of a proposed $1-for-$2 benefit offset for
its Social Security Disability Insurance (SSDI) program. Whereas under current program
rules, SSDI beneficiaries who have completed their Trial Work Period (TWP) and who earn
more than the threshold for Substantial Gainful Activity (SGA)—currently set at $1,000 per
month—are ineligible to receive benefits (that is, they lose their benefit entirely), under the
proposed benefit offset policy these individuals would retain $1 in benefits for every $2
earned above the SGA. Thus, a benefit offset may induce entry because it makes
participation more attractive for individuals who are medically eligible for SSDI benefits but
able to earn more than the SGA. The size of the population of induced entrants is a critical
input into an analysis of the effects of a benefit offset policy on overall program costs.
Although the Ticket Act included induced entry effects among the set of effects to be
evaluated with a demonstration project, SSA has determined that a demonstration project
aimed at estimating induced entry is not feasible (Tuma, 2001). As a result, SSA must now
determine an alternative method of fulfilling its mandate under the Ticket Act. In service of
that goal, this report is designed to provide SSA with two carefully selected research design
options to estimate induced entry under the proposed benefit offset policy, as well as the
information needed to evaluate each design on several dimensions: internal validity, external
validity, flexibility, economy (cost), and speed.
To accomplish these objectives, we first performed an extensive literature search and
prepared a list of candidate research designs. In January 2010, we convened a meeting with
SSA and our Technical Advisory Group (TAG), consisting of experts on the SSDI program
and research methods for estimating entry effects, where we presented this list of candidate
designs. After consultation with the TAG and SSA stakeholders in attendance, two research
designs were identified as the most promising:
•
•
a research design using stated preferences (SP)
a research design using past policy (PP) changes in a simple structural framework.
In the remainder of this summary, we highlight key findings with respect to the study
objectives.
vi
Theoretical Framework for Induced Entry
One important contribution of this report is to develop a simple theoretical framework for
understanding the mechanisms that give rise to induced entry. This framework provides a
rigorous yet intuitive starting point for an analysis of induced entry. The model is dynamic,
forward-looking, and yields steady-state conditions for SSDI claiming and employment
behavior. The model identifies several factors, such as health insurance, that figure into the
decision to apply for SSDI benefits. More importantly, the model identifies a group of
disabled, nonbeneficiary workers who would be better off claiming SSDI under the benefit
offset policy but not under the current policy, and thus could be induced to enter the SSDI
program. This group consists of individuals with earnings in a particular range defined by the
SGA threshold, the benefit offset rate, and the individual’s monthly benefit amount. This
condition plays an important role in defining the sample frame for the research design using
stated preferences.
Research Design Using Stated Preferences
A promising method for estimating potential entry effects as a result of a $1-for-$2 benefit
offset is a research design using the stated preferences method. In this method, one
administers a series of stated choice experiments designed to reveal respondents’ preferences
for claiming disability under varying program rules and economic conditions. In particular,
the SP approach consists of presenting respondents with a set of scenarios describing
different states of the world and asking them to rate, rank, or choose among different
possible actions (e.g., continuing to work versus claiming disability under varying
conditions). The scenarios are characterized by either real or hypothetical attributes (or a
mix of both), such as a benefit offset rate or earnings disregard level, and allow one to
estimate the impact of a hypothetical policy that has never been experienced by respondents.
Sampling Plan
An important feature of the SP design is that it requires new data collection. Because the
target population—potential induced entrants—has unknown characteristics and is likely
small relative to the general population, a critical issue is how to sample and screen
respondents. As noted above, economic theory offers a useful guide for winnowing down the
sampling frame to individuals in a particular range of earnings, who are most likely to make
up the target population. We identified two potential sampling frames as promising
candidates for an SP-based research design:
•
We identified the SSA administrative database of U.S. workers as an ideal sampling
frame, since it includes every worker insured for SSDI benefits in the United States
and their history of earnings and benefit receipt. Since health information is not
available in SSA’s administrative data, individuals cannot be sampled on the basis of
their likelihood of medically qualifying for the SSDI program. A health screener, such
as the 26-item screener developed by Westat in 2002 for the National Study of
Health and Activity (NSHA), could identify respondents with health conditions that
may qualify them for the program. However, because the medically eligible
vii
•
population is very small, at most one-quarter of those screened would likely be
included in the final survey.
We also identified the American Community Survey (ACS) as a potential sampling
frame. The ACS surveys roughly 1.5 million individuals ages 25–64 per year. Linking
the ACS to SSA administrative data would allow one to narrow the sample frame to
individuals with earnings in the appropriate range who are eligible for but not
receiving SSDI benefits. The advantage of the ACS is that it already includes six
questions on disability that can be used to pre-screen respondents. An additional
health screener, such as the Westat screener, could be used to further refine the
sample. However, additional research is needed to determine the fraction of
individuals likely to pass the screen.
Experimental Design
As noted above, the SP design consists of asking respondents to imagine their behavior under
a series of hypothetical scenarios. We identified three variations on the SP design that could
be used to estimate induced entry:
•
•
•
The baseline approach is the simplest approach, designed to yield an estimate of
induced entry under a $1-for-$2 benefit offset in an otherwise unchanged program
environment. This approach consists of describing the benefit offset to currently
disabled nonbeneficiaries and asking whether they would apply for SSDI benefits
under the new policy. This approach is by far the most expensive to implement, on
the order of $2.1 million, assuming a sample frame based on SSA administrative data
(excluding pilot testing and other survey-design activities). Moreover, the baseline
design does not offer any flexibility to estimate responses to variations of the benefit
offset policy.
A baseline plus approach goes a step further and specifies a statistical model for SSDI
claiming as a function of proposed program parameters (e.g., offset rate, disregard
level) and current program parameters (e.g., the SGA level), known as attributes.
This design is extremely flexible and reduces costs substantially by imposing modest
structure on the estimation problem with few additional assumptions. Additionally,
it is possible to conduct randomized choice experiments by randomly varying
hypothetical attributes over respondents, which maximizes statistical power by
setting the correlation between attributes to zero. If respondents each rate several
profiles (scenarios consisting of different attributes), then sample size can be further
reduced. Estimated implementation costs for a baseline plus design varying two
attributes, each with 3–4 levels, and asking respondents to rate 6–12 profiles range
between $381,000 and $632,000.
Finally, we propose an alternative method that achieves cost savings by eliminating
the need to screen out 75 percent of the sample based on health. We do so by
recasting health itself as an attribute to be specified explicitly in the hypothetical
scenarios presented to respondents. Introducing health as an attribute has the added
advantage of allowing one to control for health in a uniform way by designing
scenarios specifically based on SSDI medical eligibility criteria. Estimated
viii
implementation costs for an alternative design asking respondents to rate 5–15
profiles range between $204,500 and $354,000.
Summary Evaluation
The SP approach is extremely flexible and allows program parameters and other attributes of
the scenarios to be varied easily. In addition, it requires very few assumptions regarding
specification of the individual’s decision environment. The key assumption for identification
is that respondents are able to accurately forecast their behavior under new and unfamiliar
policy conditions. In addition, using health as an attribute of choice scenarios and surveying
individuals who are not medically eligible requires the additional assumption that individuals
can forecast their behavior under different health conditions. Although new data collection is
costly, SSA administrative records and the ACS provide inexpensive yet comprehensive
sample frames. In addition, presenting respondents with multiple, randomized scenarios
leads to impressive reductions in sample size without sacrificing statistical power.
Research Design Using Past Policy Changes in a Simple Structural Approach
This research design leverages past changes in the SGA threshold to estimate key behavioral
parameters that could be used to forecast entry behavior. The SGA threshold is a
fundamental program parameter, determining both initial eligibility and ongoing entitlement
to SSDI benefits, and it figures directly into the current work rules. These past policy
changes are closely related to the introduction of a proposed benefit offset, as they both
modify the shape of the budget constraint that potential entrants face. This research design
is composed of two parts: (1) a reduced form analysis of the impact of SGA changes on SSDI
applications/enrollment, which provides a potential test of whether one might expect any
induced entry under a benefit offset, and (2) a simple structural analysis, which relates the
reduced form estimates to induced entry under a specific $1-for-$2 benefit offset or a range
of offset policies.
Reduced Form Analysis
There is significant variation in real SGA levels over time, including increases, decreases, and
periods of relative stability. SSA has increased the (nominal) SGA threshold several times in
past decades, including large increases in 1990 and 1999. At the same time, inflation has led
to real declines in the SGA before and in between these increases. Since December 2000, the
SGA threshold has been indexed to a measure of annual average wages for all employees in
the United States. In addition, the SGA level is relatively more generous in areas with lower
costs of living and/or lower average wages. Therefore, there is considerable variation in the
SGA level across time and space (e.g., states or counties) when considered in relative terms.
One can construct a measure of real, relative SGA levels by dividing the Consumer Price
Index–adjusted national SGA level by a state- or county-level index of average wages. Using
SSA administrative data, one can then regress SSDI application and/or enrollment rates at
the state-year level on real, relative SGA levels along with controls for changes in
macroeconomic conditions (i.e., state and year fixed effects, and such variables as state-level
unemployment rates). Since the SGA threshold is such a fundamental program parameter—
ix
affecting both entry and ongoing entitlement—failure to detect an effect of past SGA
changes on SSDI entry might lead one to expect little or no change in entry in response to a
proposed benefit offset.
Structural Analysis
While raising the SGA threshold is not equivalent to introducing a benefit offset, both policy
changes affect approximately the same area of the budget constraint faced by potential
entrants. As a result, one could relate the SGA-induced entry effect to the benefit offset
setting using a simple structural framework. Specifically, one could specify a utility
maximization problem where individuals jointly determine their labor supply and SSDI
program participation. Once one assumes a functional form for labor supply (or,
equivalently, the indirect utility function), specifies a role for observable individual
characteristics, and assumes a distribution for unobservables, the model could be estimated
using maximum likelihood or method of moments. Once one has obtained estimates of the
utility function parameters, one could apply them to the hypothetical budget constraint
under the proposed benefit offset to simulate who would apply for SSDI under the new
program. An estimate of induced entry can then be obtained by subtracting the number of
applicants under the current policy from those under the proposed benefit offset policy. In
this framework one could make use of three sources of identification:
•
•
•
The discontinuity in the budget constraint arising from the presence of the SGA
threshold—in principle, past policy changes are not necessary to identify the
parameters of the model, as revealed preference under a nonlinear budget set in
cross-section is sufficient to identify the model, with certain assumptions (Moffitt,
1990). One assumption is that observed nonlabor income is exogenous.
Bringing sufficiently large SGA changes (across time or space) into the analysis
allows one to relax assumptions about the income elasticity of program participation
with respect to benefits. Intuitively, individuals who previously earned more than the
old SGA but less than the new SGA experience a local outward shift in their budget
constraint, as they are now eligible for SSDI benefits. The receipt of SSDI benefits
increases their total net income (earnings plus benefits) without affecting their net
wage rate (the amount they can keep as income if they work an additional hour).
This allows one to identify the income elasticity without additional assumptions
about nonlabor income. Incorporating SGA changes allows one to explicitly link the
proposed reduced form and structural analyses. Specifically, one could estimate the
model using a method of moments strategy and include the estimated reduced form
effect as a moment to be matched.
Finally, the induced entry project is only part of a portfolio of projects funded by
SSA to estimate potential impacts of a benefit offset. Another such project is the
Benefit Offset National Demonstration (BOND) project, which is investigating the
effect of the benefit offset on the labor supply of current beneficiaries. The BOND
gives one the opportunity to observe actual responses to the exact change in budget
constraint under the benefit offset. A critical drawback of the BOND for the purpose
of estimating induced entry is that it only provides information on labor supply; that
x
is, by design, it does not include information on the participation decision. However,
the BOND does provide a valuable opportunity to incorporate additional variation
into the estimation (e.g., by including an additional moment to be matched) or to
test the ability of the model to make out-of-sample predictions.
Summary Evaluation
This research design is fast and inexpensive. It utilizes data on the group most likely to
approximate marginal entrants under a benefit offset, actual SSDI applicants. In addition, it
provides a fair amount of flexibility in that it would be easy to modify the budget constraint
in the decision problem used to simulate behavior under hypothetical rules. A drawback of
the approach is that it relies heavily on distributional and functional form assumptions.
However, opportunities abound for testing and relaxing some of these assumptions by
exploiting the “natural experiments” arising from past SGA changes as well as an actual
randomized experiment, the BOND project.
Both research designs were determined to be capable of providing SSA with credible
estimates of induced entry into SSDI resulting from a benefit offset with relatively small sets
of assumptions. In addition, both approaches offer a great deal of flexibility and allow for a
range of estimates that would provide valuable insight into how potential SSDI applicants
make decisions regarding program participation. While both research designs produce
partial-equilibrium “steady-state” estimates of induced entry, they yield parameter estimates
that could be used to forecast entry over time, accounting for changing economic and
demographic conditions (e.g., trends in population aging, health, and labor demand). In a
head-to-head comparison, there is no clear winner, as both research designs are strongest on
different criteria. Whereas the SP design may offer slightly greater flexibility and require
fewer and weaker assumptions, the PP design is cheaper, faster, and uses data on individuals
who most closely approximate marginal entrants. As SSA has stressed a strong desire for a
range of plausible induced entry estimates, one promising avenue for further research is to
implement both research designs and compare the results.
xi